This study used a cross-sectional design to gather stroke patients who were admitted to Shanghai East Hospital between January 2020 to August 2022. It comprised patients (18 years of age or older) whose diagnosis of stroke was verified by cranial CT or MRI within 24 h of admission.

Patients without a reported Barthel Index score at admission or those whose serum albumin levels were not tracked were excluded. Figure 1 illustrates the patient selection process. The cross-sectional study was reported concerning the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines and approved by the Shanghai East Hospital Ethics Committee (No. 2024055). As a retrospective cross-sectional study with anonymous data collection, informed consent was not required. All procedures and methods followed the World Medical Association’s Declaration of Helsinki on Human Experimentation ethics.

Participants´ demographic and lifestyle data was gathered. Individual anthropometric information was entered into the hospital’s electronic medical record system, including height, weight, sex, and age. In summary, these anthropometric measurements were taken at the time of the patient’s hospital admission during their physical examinations. Weight and height assessments followed the World Health Organization’s guidelines. Weight (kg) divided by height squared (m2) yielded the body mass index (BMI). There were two categories for smoking status: “current or ever smoking” and “current or ever drinking” for drinking status. A history of documented hypertension or blood pressure of at least two instances of ≥140 mmHg (systolic) or ≥ 90 mmHg (diastolic) following the acute phase of a stroke was used to diagnose hypertension. Information on diabetes, coronary heart disease, atrial fibrillation, cancer, and previous stroke history was extracted from the electronic medical records.

The National Institute of Health Stroke Scale (NIHSS) was used to evaluate the severity of the stroke at the time of admission (16, 17). The TOAST criteria classified ischemic strokes into five subtypes: large-artery atherosclerosis, small-vessel occlusion, cardioembolism, stroke with another identified etiology, and stroke with unknown etiology (18). Additionally, Serum albumin level was measured using a Roche analyser within 24 h of the onset of stroke. Serum albumin levels are measured using the bromocresol green (BCG) method, based on the principle that albumin specifically binds to the dye bromocresol green under acidic conditions, resulting in a color change that is quantified spectrophotometrically. The method is widely used and economical; when obtaining serum samples from patients, ensure that they are free from hemolytic, lipemic, or jaundiced interference to ensure the accuracy of the measurement.

Activities of daily living (ADL) comprise the basic actions that involve caring for one’s self and body, including personal care, mobility, and eating. The Barthel Index (BI) is a widely used scale for assessing a patient’s ability to perform basic activities of daily living (ADLs). It was first developed in 1965 by Dorothea Barthel and Florence Mahoney in the United States (19). The index evaluates ten functional areas, including feeding, bathing, dressing, toileting, mobility, walking, stair climbing, and bowel and bladder control (20, 21). The total score ranges from 0 to 100, with higher scores indicating greater independence in daily activities. The Barthel Index is extensively used to assess functional status in various conditions, particularly in the study of Stroke (22), Spinal Cord Injury (23), Dementia (24), and Parkinson’s Disease (25). Several previous studies in China have also used BI to assess ADL capacity in stroke patients (26, 27).

Previous studies have further categorized subjects into two groups based on activities of daily living (ADL): those with a high ADL score of 40 or more and those with a low ADL score of 40 or less (28). When any patient’s Barthel score is less than 40, it means that neither their mobility abilities nor ability to feed themselves, groom themselves, or control their sphincters are autonomous (29). The patient cannot perform daily tasks independently and relies on assistance from others. This level of impairment is defined as severe impairment of activities of daily living.

Every patient was the subject of a descriptive analysis. Numbers (percentages) were used to express categorical data, and depending on the distribution, continuous data were shown as the median (interquartile range) or the mean ± standard deviation. To evaluate differences among groups, one-way analysis of variance (for normally distributed data), Kruskal–Wallis tests (for non-normally distributed data), and chi-square tests (for categorical variables) were performed. The odds ratios (OR) and 95% confidence intervals (CIs) for the association between serum albumin and severe impairment of ADL in stroke patients were calculated using logistic regression models. Model 1 was calibrated for BMI, age, sex, drinking, and smoking status. Age, sex, BMI, drinking and smoking status, hypertension, diabetes, cancer, atrial fibrillation, coronary heart disease, and history of stroke were all taken into account while adjusting Model 2. All of the variables from Model 2 were included in Model 3, along with TOAST classification and NIHSS.

Additionally, after controlling for the factors in Model 3, we used restricted cubic spline (RCS) regression to analyze curvilinearity and investigate the dose–response association between serum albumin and severe ADL impairment. To evaluate the association threshold between serum albumin and severe ADL impairment after stroke, we employed a smoothed binary logistic regression model. Likelihood ratio tests and bootstrap regression were used to identify significant inflection points in this relationship.

Furthermore, we investigated several variables that could alter the association between serum albumin and severe ADL impairment following a stroke. The variables that were analyzed were sex; age (less than 60 years old versus more than 60 years old); BMI (less than 25, 25–29.9, versus more than 30 kg/m²); smoking and drinking status (yes or no); hypertension (yes or no); diabetes (yes or no); and TOAST classification (large-artery atherosclerosis, small-vessel occlusion, cardioembolism, stroke with another identified etiology, and stroke with unknown etiology).

Using multivariate logistic regression, we assessed subgroup heterogeneity, and the likelihood ratio test examine interactions between subgroups and serum albumin.

To examine the robustness of the results, we did a sensitivity analysis by interpolating the NIHSS, which had a high number of missing data, using median values. No, a priori statistical power analyses were carried out because the sample size was established exclusively using the data that were available data. R version 4.3.1¹ and Free Statistics version 1.8 were used for all analyses. A descriptive study was carried out on each individual. A two-tailed p-value of less than 0.05 was specified as statistically significant.

The study initially included 3,365 stroke patients. Among these, 73 patients were eliminated due to missing data on ADL, whereas 899 patients were omitted due to missing albumin levels. As a result, this cross-sectional study comprised 2,393 patients from Shanghai East Hospital from 2020 to 2022. Figure 1 provides a detailed illustration of the selection process, depicting both inclusion and exclusion criteria.

Table 1 summarizes all subjects´ baseline characteristics, classified by serum albumin quartiles.

The average age of the patients was 69.6 ± 11.6 years, with 1,538 (64.3%) being male. Patients with higher serum albumin levels were younger, male, had a slightly higher BMI, higher BI scores, lower NIHSS scores, had a history of hypertension, no history of coronary heart disease and atrial fibrillation, were non-smokers, and had atherosclerosis according to the TOAST classification. The four groups had significant differences in sex, age, BMI, Barthel Index Score, NIHSS, Hypertension, Coronary heart disease, Atrial Fibrillation, Smoking Status, and TOAST (all p < 0.05).

After adjusting for confounding factors, the multivariate analysis revealed a 7% decrease in severe impairment of ADL after stroke for every unit (g/L) increase in serum albumin levels. When blood albumin levels were analyzed using quartiles, a negative relationship was found between serum albumin levels and severe impairment of ADL after controlling for relevant variables. In comparison to patients with lower levels of serum albumin (<37.4 g/L), the adjusted odds ratio (OR) for severe impairment of ADL after stroke for those in the second quartile (Q2: 37.4–40.21 g/L), third quartile (Q3: 40.21–42.80 g/L), and fourth quartile (Q4: >42.8 g/L) were 0.68 (95% confidence interval [CI]: 0.40–1.15, p = 0.148), 0.55 (95% CI: 0.32–0.97, p = 0.04), and 0.65 (95% CI: 0.37–1.15, p = 0.139), respectively (Table 2).

In Figure 2, the link between serum albumin and severe impairment of ADLs after stroke displayed an L-shaped curve (nonlinear, p = 0.002). The link between serum albumin levels and severe impairment of ADLs among stroke patients shows an inflection point around 38.0 g/L. In the threshold analysis, the odds ratio (OR) for severe impairment of ADLs among stroke patients with serum albumin levels less than 38.0 g/L was 0.680 (95% CI: 0.568–0.814, p < 0.001). When serum albumin levels were greater than or equal to 38.0 g/L, the severe impairment of ADLs no longer decreased with rising serum albumin levels, indicating that the threshold had been reached (Table 3).

Figure 3 illustrates the relationship between serum albumin levels and severe limitations in ADL, as analyzed across various subgroups. Serum albumin levels were associated with severely impaired ability to perform activities of daily living in male (OR, 0.92; 95% CI, 0.88–0.96), female (OR, 0.86; 95% CI, 0.81–0.92), more than 60 years (OR, 0.91; 95% CI, 0.87–0.94), with BMI less than 25 kg/cm² (OR, 0.9; 95% CI, 0.85–0.95), BMI 25–29.9 kg/cm² (OR, 0.91; 95% CI, 0.85–0.97), smoking (OR, 0.91; 95% CI, 0.86–0.96), non-smoking (OR, 0.89; 95% CI, 0.85–0.94), alcohol (OR, 0.88; 95% CI, 0.8–0.96), no alcohol (OR, 0.9; 95% CI, 0.86–0.94), hypertension (OR, 0.89; 95% CI, 0.86–0.93), diabetes (OR, 0.91; 95% CI, 0.85–0.97), no diabetes (OR, 0.92; 95% CI, 0.88–0.97), and those with TOAST classification large-artery atherosclerosis (OR, 0.91; 95% CI, 0.86–0.96), small-artery occlusion (OR, 0.89; 95% CI, 0.82–0.95), stroke of unknown etiology (OR, 0.81; 95% CI, 0.7–0.94). There was no significantly association among aged less than 60 years, BMI more than 30 kg/cm², or those without hypertension or in the Cardioembolism and Stroke of another determined etiology subgroups. After subgroup analysis according to sex, age, BMI, smoking status, alcohol status, hypertension, and diabetes, TOAST classification, no significant interactions were found in any subgroup (all p-values for interaction >0.05).

Due to the high rate of missing NIHSS and TOAST data, we found a more robust association between serum albumin and severely impaired ADL of stroke after the multiple interpolation of NIHSS and TOAST. After adjusting for confounding factors, multivariate analysis showed that for every 1 g/L increase in serum albumin levels, there was a 7% reduction in severe ADL impairment post-stroke. Compared with those with lower serum albumin levels (<37.4 g/L), the adjusted ORs for serum albumin and ADL in Q2 (37.4–40.21 g/L), Q3 (40.21–42.80 g/L), and Q4 (>42.8 g/L) were 0.57 (95% CI: 0.38–0.86, p = 0.007), 0.50 (95% CI: 0.32–0.77, p = 0.002), and 0.52 (95% CI: 0.33–0.81, p = 0.004), respectively (Supplementary Table S1).